[blyatman]

I have two define_on_demand macros called "run1" and "run2".

"run1" spits out an array x. "run2" uses x, as well as a parameter y obtained from Get_Input_Parameter, to spit out another output z, so z = z(x,y).

I am trying to build a workflow that does something like this:
1. Call "run1" to get x.
2. Set a value of y, call "run2" to get z, save results.
3. Repeat previous step for different values of y.

The reason I do not want to combine "run1" and "run2" into a single define_on_demand macro is because I only need to execute "run1" once. If I combine it into a single macro with "run2", then "run1" would execute every single time with "run2", which is unnecessary and potentially time-consuming.

Is there a way to do this? I have considered using scheme variables, but there is no option to define an array using a scheme variable.

[Siba11]

Hi. You say that you need to execute "run1" only once. This means that while the parameter 'y' will attain different values, the array will be treated as a constant. So why don't you add that array as an input variable in your second UDF (run2). And then execute only "run2" for as many times as you want.

This will save you from having to write any scheme scripts.

[blyatman]

Originally Posted by Siba11

Hi. You say that you need to execute "run1" only once. This means that while the parameter 'y' will attain different values, the array will be treated as a constant. So why don't you add that array as an input variable in your second UDF (run2). And then execute only "run2" for as many times as you want.

This will save you from having to write any scheme scripts.

When you say to add that array into run2, do you mean hardcoding 'x' into run2? I would be using this workflow for different cases, so hardcoding 'x' would not be practical.

[Siba11]

Hi. Maybe I'm misunderstanding. What is this "x" array exactly? Like, is it an array storing the coordinates of all mesh cells or something like that?

As you might be knowing, DEFINE_ON_DEMAND doesn't execute every iteration as it is outside the solver loop. So whatever value this "run1" is spitting out, it must be independent of the iteration number. So, you can execute "run1" once, get the array "x", and hardcode it into "run2".

(UNLESS of course 'x' is something that's going to change on the fly.)

[blyatman]

Originally Posted by Siba11

Hi. Maybe I'm misunderstanding. What is this "x" array exactly? Like, is it an array storing the coordinates of all mesh cells or something like that?

As you might be knowing, DEFINE_ON_DEMAND doesn't execute every iteration as it is outside the solver loop. So whatever value this "run1" is spitting out, it must be independent of the iteration number. So, you can execute "run1" once, get the array "x", and hardcode it into "run2".

(UNLESS of course 'x' is something that's going to change on the fly.)

'x' is going to be an array of arrays, and it will be a mesh-related quantity that is independent of the solution. This array could have a million elements, and for each element, there could be a few thousand elements. As a result, it is impractical to hardcode it, which is why I am trying to figure out a way to pass it into "run2" (not to mention that I will be running a script that does this for other meshes, so hardcoding 'x' would again be out of the question). Hope this clears up the confusion.